[om-list] Coin Tossing Theory
Mark Butler
butlerm at middle.net
Fri Aug 29 15:22:43 EDT 2003
Suppose that you, having no prior knowledge of coins, have someone come
up to you and say through experimental research he has determined that
tossed coins always end up heads.
Having a skeptical mind and a curious bent, you decide to test his
theory through experiment. Of course, on average the coin ends up heads
only about half the time. There are a couple of ways you might treat
this data:
Conventional Boolean logic:
Each time the coin toss is heads, it adds a small amount of
confirming evidence for the hypothesis. The first time the coin
toss is tails, that adds a sufficient amount of disconfirming
evidence to disprove the theory. You remark on the correctness of
the maxim, "All generalizations are false". You are convinced that
the theory has no likelihood of being right. Your degree of belief
in his theory is "absolutely false".
Fuzzy logic:
You tally up the net number of heads and the net number of tails,
divide, get a number near 1/2 and conclude, "I am confident he is
half right, after all he could have been wrong all the time. Your
degree of belief in his theory remains neutral on a scale ranging
between "always wrong" and "always right".
Given the presence of minor errors in most theories, I submit that fuzzy
logic is much more appropriate for estimating truth values of
propositions in a semantic network than traditional Boolean logic. With
the latter, we have to toss out a theory completely on the first
inconsistency. It is much more useful to track the predictive value of
propositions, that like all generalizations admit of exceptions, as a
continuous variable, even when the subject data is boolean in nature, as
in the example.
It drives me crazy to hear people claim that Quantum Mechanics implies
that Newton's Laws are now "false", when in actuality they are extremely
accurate on a very wide scale of applicability.
- Mark
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://six.pairlist.net/pipermail/om-list/attachments/20030829/45a1d30f/attachment.html
More information about the om-list
mailing list